Projects
MLOps for Federated Learning
Undergraduate final project: heart disease classification with Flower and MLflow. Implemented experiment tracking, best-model registration, and centralized container logs with Docker, Grafana Loki, and Alloy.
Karsakito AI Backend Engine
Built FastAPI services for Minangkabau translation, writing, and language etiquette. Integrated LLM providers through LiteLLM, Redis caching, input guardrails, and Langfuse tracing for latency and model usage.
Klop! AI Assessment Backend
Hackathon project: developed APIs for generating technical assessments and scoring answers with Qwen and Kolosal AI. Deployed the backend to a DigitalOcean Droplet with Cloudflare routing and Caddy HTTPS.
Order Processing Microservices
Learning project connecting order, payment, and inventory services through Go, gRPC, and RabbitMQ. Explored event-driven workflows, message retries, dead-letter queues, and separate PostgreSQL databases with Docker Compose.
ML Training, Serving & Monitoring
Course project exploring the ML lifecycle through model experiments, a CI workflow, API serving, and monitoring. Combined FastAPI and Docker with Prometheus and Grafana to practice operating an ML service.
Simobi+ Review Sentiment Analysis
Processed app reviews and compared SVM, Bi-LSTM, and IndoBERT for sentiment classification. Used manual annotation and pseudo-labeling to explore learning from limited labeled data, with separate preprocessing and evaluation notebooks.